Win My Argument

A political debate comparison tool that generates arguments from different political viewpoints on any given topic, displayed in an interactive side-by-side TUI format.

Description

Win My Argument is a Python TUI application that simulates political debates between left-wing and right-wing perspectives. It uses pydantic-ai agents with local LLM models (via Ollama) to generate realistic arguments that compare different political viewpoints.

The application:

1. Takes a debate topic as input (custom or from predefined list)
2. Generates responses from both right-wing and left-wing perspectives in parallel
3. Uses DuckDuckGo search to gather relevant information for each perspective
4. Displays arguments side-by-side with model performance statistics
5. Allows comparison of arguments in an intuitive TUI layout

Screenshot

Installation

# Clone the repository
git clone https://github.com/c2p-cmd/win_my_argument.git
cd win_my_argument

# Create a virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

Usage

Run the interactive TUI application:

python main.py

Interactive Features

1. Welcome Screen - Application introduction with key information
2. Topic Selection - Choose between:
   • Typing your own debate topic
   • Selecting from 10 predefined political topics
3. Live Debate Generation - Watch as arguments are generated
4. Side-by-Side Comparison - View arguments in split-pane layout:
   • 🔴 Right-wing argument (red, left panel)
   • 🔵 Left-wing argument (cyan, right panel)
5. Performance Statistics - See detailed model stats for each side:
   • Total tokens used
   • Input/output token breakdown
   • Tokens per second (generation speed)
   • Response time

Application Flow

graph TD
    A["🚀 Start Application"] --> B["Display Welcome Screen"]
    B --> C{"Select Input<br/>Method"}
    C -->|Option 1| D["Type Custom Topic"]
    C -->|Option 2| E["Choose from<br/>Predefined Topics"]
    D --> F["Debate Topic<br/>Selected"]
    E --> F
    
    F --> G["Search Information<br/>DuckDuckGo"]
    G --> H["Right-Wing Agent<br/>Generates Arguments"]
    G --> I["Left-Wing Agent<br/>Generates Arguments"]
    
    H --> J["AI Model<br/>Processes Response"]
    I --> J
    
    J --> K["Display Side-by-Side<br/>Comparison with Stats"]
    
    K --> L{"Continue<br/>Debate?"}
    L -->|Yes| C
    L -->|No| M["🎉 Exit Application"]
    
    style A fill:#00ff00
    style M fill:#ff0000
    style H fill:#ff6b6b
    style I fill:#4ecdc4
    style K fill:#ffd93d

Loading

Project Structure

win_my_argument/
├── main.py                    # Main TUI entry point
├── debater/                   # Core debate module
│   ├── __init__.py
│   └── debate/
│       ├── __init__.py
│       ├── agents.py          # AI agent configuration
│       ├── model.py           # Data models and system prompts
│       └── service.py         # Debate orchestration and stats
├── requirements.txt           # Python dependencies
├── README.md                  # This file
└── LICENSE                    # MIT License

Configuration

The application defaults to US political figures, but can be configured for other countries by modifying the __country variable in debater/debate/model.py.

Currently supported countries

• US - Donald Trump vs Joe Biden
• DE - Alice Weidel vs Olaf Scholz
• IN - BJP vs INC

To change country, edit this line in debater/debate/model.py:

__country = "US"  # Change to "DE" or "IN"

Requirements

The application requires:

• Python 3.8+
• Ollama - For running local LLM models

Key Python dependencies

• pydantic-ai - Agent-based AI workflows
• rich - Beautiful terminal UI components
• prompt_toolkit - Interactive command-line interface
• duckduckgo_search - Information retrieval
• dotenv - Environment configuration

Features

• ✨ Interactive TUI - Beautiful terminal interface with Rich
• ⚖️ Side-by-Side Comparison - Easy argument comparison
• 📊 Performance Metrics - Track model efficiency with tokens/sec
• 🎯 Predefined Topics - 10 pre-selected debate topics
• 🔧 Customizable - Support for multiple countries and viewpoints
• 🚀 Fast Generation - Uses local Ollama models for privacy

How It Works

1. User Input - Select or enter a debate topic
2. Parallel Processing - Both agents process the topic simultaneously
3. Information Gathering - DuckDuckGo search for context
4. Argument Generation - LLM generates arguments matching political perspective
5. Display - Results shown in split-pane with statistics

License

MIT